ECHOMOP News

Martin Clayton, Starlink. 24th October 1996.

This HELP facility accompanies ECHOMOP V3.2-0.

This is an interim release of ECHOMOP with many bug fixes and changes.

Digital UNIX users will notice a significant performance improvement from v3.0.

ECHOMOP is currently being fully supported by Starlink. You are encouraged to make suggestions for improvements and report bugs. Bug fixes will be made available in further releases as they become available.

General notes

• Several modifications have been made to enhance the speed for the program. Generally a reduction of about 35% in execution time can be expected.

New features

• A new task ech_genflat has been added. This task outputs flat-field balance factors as generated by ech_ffield to an image. The image can then be inspected using, e.g., KAPPA DISPLAY. For more details see the on-line HELP or SUN/152.
• The design of the interface in parameter editors has been altered to make them a little easier to use.
• The parameter TUNE_USE_NXF has been "enhanced". Previously, this parameter set the fraction of an order (in the dispersion direction) to be used when profiling; setting the parameter to a value of 1.0 selected a "special" mode where each order was separately profiled. This behaviour remains the same; however, it is now possible to select individual-order profiling and set the fraction of each order to be used. For example, a value for TUNE_USE_NXF of 1.2 selects the central 20% of each order and individual-order profiling.
• 'QUIT' and 'Q' have been added as aliases for EXIT in the echmenu top-level menu.
• 'Q' has been added as an alias for 'E' (EXIT) in the ech_plot top-level menu.
• The task ech_fcheck now checks both input and trace frames for bad-pixel values in their data arrays. The trace frame is now checked for saturated pixels.
• In the plotting task the full option menu is displayed only once. The menu can be redisplayed by pressing 'M' as in other tasks.
• In the plotting task when displaying reduction data the default prompt now automatically sets its self to point at the next order for the last data plotted. For example a plot of 'OBJ[1,1]' - the first order of the object - will set the default value to 'OBJ[1,2]' - the next order. This means it is much faster to check through all the orders.

Documentation

• ECHMENU option 1.5 has been documented in SUN/152.
• Some errors in SUN/152 have been fixed, the text clarified in a few places, and several missing parameter details added.
• Some errors in the on-line and hypertext HELP texts have been corrected.
• The out-of-date and uninformative "Inputs-Outputs" and "Method" entries in the on-line and hypertext HELP text have been removed.

Parameter default changes

• The default value of TUNE_MXSMP has been changed from 200 to 500. This reflects the increase in size of CCDs in recent years.
• The advertised default value of TUNE_TWTHR (0.9) was incorrect and has been corrected to 0.95.
• The advertised default value of TUNE_AUTLOC (TRUE) was incorrect and has been corrected to FALSE.
• The advertised default value of TUNE_NPOLY (7) was incorrect and has been corrected to 4.
• The default value of TUNE_ARCHIVE has been changed to FALSE.
• The default value of MIN_DISPERSION has been changed to 0.01 which better reflects the dispersions used in echelle instruments.

Other changes

• ECHOMOP no longer uses the NAG library.
• The module definition database has been removed. Modules are defined as the program starts up.
• Only one call to PSX_UNAME is made per invocation of the monolith. Previously, several routines called PSX_UNAME.
• Calls to PGPOINT have been changed to call PGPT.
• The subroutine calculating median values has been enhanced.
• There is only one routine for mean/median/most-common-value calculation.
• Several table look-up routines have been modified to exit-on-match, rather than searching the full table for an entry.
• The trace-clipping task menu is only displayed once which speeds up the process of trace clipping. The full menu can be displayed using option "M" in the same style as the profiling tasks.
• In the spatial-profiling task the display of "below"-trace pixel-distances has been changed to reflect the syntax used. For example the dekker edge is now said to be "20" pixels "below trace" rather than "-20" pixels below which suggests that it is above the trace.
• Messages of the form "Parameter XXX is set to non-default value: nn" have been shortened.
• Some of the main menu text has been changed.
• Several (140) calls to ECH_SET_CONTEXT which had no effect have been removed from subroutines. This may lead to a small speed-up for some operations.
• When DISPLAY=TRUE overlaid trace plots are labelled with the relevant order number. The display scaling is calculated only once per frame, rather than once per order, which speeds up this option.
• The routine ECH_FIND_CENTRE has been modified, speeding up centre-of-gravity mode tracing and echmenu option 9 (locate arc line candidates).
• The file ech_dynamix_index_index.f (which was not used) has been removed.
• The file ech_makefits.f (which was not used and is not needed) has been removed.
• The file ech_plot_id_lines.f (which was not used) has been removed.
• In the dekker/object-profiling task (echmenu option 4) graph titles now include the number of the order displayed.
• Many FORMAT statements have been changed by the addition of "1P" scale factors. This makes output numbers easier to read.
• The files ech_loop_nvariable.f and ech_loop_variable.f have been removed due to the process of speeding up the main ECHOMOP routine and introduction of extra functions.
• The IDX_ parameters were being typed as _REAL by the program whilst they were (correctly) declared as _INTEGER in interface files. This has been corrected.
• In the order-blaze fitter the plots of function-verses-fit now use the same colour scheme as in the order-trace fitter.
• The displays and menus in the interactive flat-field modeller have all been tidied up.
• In the trace-clipping task the '.' option now deletes the nearest point to the cursor as advertised. Previously this deleted the nearest point in X. This change also affects the blaze fitter and sky modeller.
• ECH_FATAL_ERROR and ECH_CALC_TRACE have been modified to speed up the program.

Fixes

• In the order-fitting/clipping process it is now possible to select an order of fit less than 4.
• In the ech_spatial task it is no longer possible to set the value of DEK_BELOW greater than DEK_ABOVE and vice versa. Previously this would cause various problems.
• Internal arrays in profiling tasks have been enlarged and range checks are now performed prior to array usage. This has removed some intermittent crashes.
• Most format statements using Hollerith characters have been updated.
• An error in the data access layer relating to default object dimensions has been fixed.
• Previously, in plots of order-trace versus fitted-curve the points of the order trace were shifted down. This gave the impression that the fit was offset by a constant. This bug has been removed. This bug also affected the order trace overlaid on image plots.
• A limitation in ECH_MEAN_MEDIAN whereby only the first 5000 values of a dataset would be used in median determination has been removed.
• An error causing a crash for USE_MEDIAN=TRUE in the order-location task has been removed. USE_MEDIAN=TRUE now works OK.
• An error in the call sequence causing unaligned memory access and crashing ECH_DECOS2 has been removed.
• Spurious "Unknown fitting function" messages appearing when using ECH_DECOS2 have been removed.
• Centroid-mode tracing now includes a filter for bad values in images. Previously, these bad values would break the centroiding code causing a crash with a floating-point exception.
• The spatial-profiling task now includes various filters for bad values which would break it.
• A commonly-occurring divide-by-zero error in the optimal extraction algorithm implementation has been removed.
• The simple extraction algorithm will now handle bad-pixel values as well as, or instead of, quality arrays in object images.
• The simple extraction algorithm will now handle bad-pixel values in arc images.
• The profile-weighted extraction algorithm will now handle bad-pixel values as well as, or instead of, quality arrays in object images.
• The profile-weighted extraction algorithm will now handle bad-pixel values in arc images.
• The HELP facility is now accessible from the echmenu program (bug in v3.1-0).
• The quick-look extraction algorithm will now handle bad-pixel values as well as, or instead of, quality arrays in object images.
• The quick-look extraction algorithm will now handle bad-pixel values in arc images.
• After running the quick-look extraction the colour of plots is now reset to black rather than being left blue.
• The 2-D simple-extraction algorithm will now handle bad-pixel values in arc images.
• The 2-D simple-extraction algorithm will now handle bad-pixel values as well as, or instead of, quality arrays in object images.
• The arc-line width estimator routine can now handle bad values in the image. The routine has been improved to scale data according to the number of good values obtained, previously no scaling was applied.
• After running the trace consistency checker the colour of plots is now reset to black rather than being left red.
• The blaze-fitting option now checks for the case TUNE_NOFLAT=TRUE, previously it would simply crash in this case.
• After running the order trace plotting task the colour of plots is now reset to black rather than being left red.
• Image displays autoscale to suit the data rather than using fixed (incorrect) values dependant on the image dimensions.
• Order tracing with DISPLAY=TRUE now works, previously this caused a crash due to a type mismatch in a subroutine call.
• Previously, order tracing with fitted traces displayed overlaid on the traced image only worked when some other plot had been made in the session. This can now be the first plot in a session.
• A crash when running the trace consistency checker with DISPLAY=TRUE has been removed.
• Evaluation of SPLINE-fitted order traces in the order-tracing task was not done correctly, causing the program to crash in some cases. This bug has been removed.
• Reporting of messages relating to NULL (!) and ABORT (!!) responses to parameter prompts is no longer deferred.
• The faulty arc-line database file $ARCDIRS/THAR.sdf has been replaced.
• Incomplete database entries in $ARCDIRS/ THAR.sdf and CUAR.sdf have been filled. The first 10 and last 11 entries, which should be partially present, were completely omitted. The database builder has been corrected to add the partial entries.
• When using ech_idwave, in the case of one identified feature in an order, the program would crash when trying to determine the wavelength range for the order. This bug has been removed.
• Option 'M' in the ech_idwave task, order-processing menu, no longer crashes the program.
• In the ech_idwave task, in the event of a wavelength polynomial being of such a low order that all points are fitted exactly (giving zero RMS error) the program no longer crashes on a divide-by-zero.
• The interactive mode of ech_idwave, options '<' and '>' no-longer cause a crash when the display is not zoomed.
• In the interactive mode of ech_idwave, option 'I' would work only once per option 'P' this has been fixed.
• In the interactive mode of ech_idwave, option 'I', the list of nearby features is centred on the selected feature, rather than the next feature up in wavelength.
• Arc line database files no longer have to have write access enabled. This means that files in $ARCDIRS do not have to be copied to an ECHOMOP user's working directory.
• Several missing parameters have been added to the interface files.
• An infinite loop on detection of bad columns or rows in ech_fcheck has been removed.
• A divide-by-zero error in ech_ffield for local mean or median calculation has been removed.
• In ech_ffield the calculation of local medians now uses the correct start point in the X-direction. Previously, the median for the start point X-TUNE_FFLSMP rather than X-(TUNE_PFLSMP/2), was being used.
• In ech_fitblz, a floating point overflow could occur when automatic fitting became unstable. The overflows are now clipped.
• The type of the parameter TUNE_SCFRACT was incorrectly set as _INTEGER. It is now _REAL.
• Several internal changes to the scrunching routines have been made to prevent the scrunched spectra being reflected in the wavelength axis, i.e., all the fluxes being negative.
• Errors in the sky modeller which would lead to attempted processing of points outside the bounds of an image have been removed.
• The sky modeller now filters out bad-pixel values.
• The object profile modeller now filters out bad-pixel values.
• In several tasks, notably the extractions, weighting of the contribution from "boundary" pixels has been corrected to a smooth function. The function previously had two discontinuities at the top and bottom edges of the extraction channel - introducing a "jump" in some data.
• Errors in the pixel-weighting scheme implementation and pixel selection in the blaze fitter have been removed.
• ECHMENU Option 1.2 (ech_decos1) is now allowed to be an automated step. Previously the value of TUNE_CRTRC was always taken as FALSE in this case. In the same way, Options 11.3 and 11.4 now also work as elements in a TUNE_AUTOMATE request.
• NDF_SQMF has been used to switch off automatic QUALITY component checking as ECHOMOP does this itself.
• The copy-last-plot-to-hardcopy device option in ech_plotter now works.
• Several divide by zero opportunities in the 2-D sky modeller have been removed.
• The 2-D sky modeller now handles bad-pixel values in input data.
• The routine which identifies order numbers now works, this helps automatic wavelength calibration proceed faster.